NetBSD Problem Report #58594

From www@netbsd.org  Wed Aug 14 02:46:20 2024
Return-Path: <www@netbsd.org>
Received: from mail.netbsd.org (mail.netbsd.org [199.233.217.200])
	(using TLSv1.3 with cipher TLS_AES_256_GCM_SHA384 (256/256 bits)
	 key-exchange X25519 server-signature RSA-PSS (2048 bits) server-digest SHA256
	 client-signature RSA-PSS (2048 bits) client-digest SHA256)
	(Client CN "mail.NetBSD.org", Issuer "mail.NetBSD.org CA" (not verified))
	by mollari.NetBSD.org (Postfix) with ESMTPS id 279791A9242
	for <gnats-bugs@gnats.NetBSD.org>; Wed, 14 Aug 2024 02:46:20 +0000 (UTC)
Message-Id: <20240814024619.18C621A9243@mollari.NetBSD.org>
Date: Wed, 14 Aug 2024 02:46:19 +0000 (UTC)
From: campbell+netbsd@mumble.net
Reply-To: campbell+netbsd@mumble.net
To: gnats-bugs@NetBSD.org
Subject: Kernel FIOASYNC/FIOSETOWN/FIOGETOWN protocol is incoherent
X-Send-Pr-Version: www-1.0

>Number:         58594
>Category:       kern
>Synopsis:       Kernel FIOASYNC/FIOSETOWN/FIOGETOWN protocol is incoherent
>Confidential:   no
>Severity:       serious
>Priority:       medium
>Responsible:    kern-bug-people
>State:          open
>Class:          sw-bug
>Submitter-Id:   net
>Arrival-Date:   Wed Aug 14 02:50:00 +0000 2024
>Originator:     Taylor R Campbell
>Release:        current, 10, 9, ...
>Organization:
The NetBSD SIGIOdation
>Environment:
>Description:
Various drivers implement FIOASYNC/FIOSETOWN/FIOGETOWN in their own ways.  The kernel provides fsetown/fgetown/fownsignal(9), but the usage rules aren't clear enough:

1. Is the caller required to serialize these?  (Probable answer: yes)

2. Can these be used in interrupt context (or under a spin lock), or in soft interrupt context?  (Probable answer: no, but bpf(4) does try to use fownsignal(9) like that.)

3. What happens if the process terminates and the pid/pgid is recycled?  (Probable answer: nothing good; I don't see anywhere the kernel might clean up the references, in case the file in question is held by a different process.)
>How-To-Repeat:
code inspection
>Fix:
1. Clarify the man page.
2. Figure out how to deal with cleanup.  Maybe provide an opaque `struct fown' data structure for callers to wire up the weak references for concurrent teardown on process termination.
3. Audit all the callers.

Home	PR Database Search